phisco (@phisc0) header
phisco (@phisc0) profile picture

Staff software engineer: @upbound_io | maintainer: @crossplane_io | ex: @controlplaneio, @cloudnativepg @ @EDBPostgres, @sighup_

Milano, Lombardia
Joined November 2009
Photo by phisco (@phisc0)Photo by phisco (@phisc0)Photo by phisco (@phisc0)Photo by phisco (@phisc0)Photo by phisco (@phisc0)Photo by phisco (@phisc0)Photo by phisco (@phisc0)Photo by phisco (@phisc0)Photo by phisco (@phisc0)Photo by phisco (@phisc0)
Fork your dependencies, trim them to only your use case, never update unless it breaks for your users. I’ve been vocal about this for 10+ years. I’ve always said that updating is way riskier than latent bugs (which can be tracked and CVEs monitored). If you are updating a dependency, it’s on you to analyze every single commit in the full transitive set of dependencies. If you dont see anything compelling, dont update! I remember at HashiCorp once in awhile an engineer would try to update a dep or replace a DIY lib with an external one and id always ask “show me the commit we need.” Dont update for the sake of it. Feeling pretty swell about this mentality with all the supply chain attacks happening.
292
784
243
9,124
1,284,633
software is dead. sad.
171
66
77
1,373
211,784
Periodic reminder: Good security architecture minimizes the number of bugs and CVEs you have to care about.
9
45
10
233
18,129
phisco retweeted
The actual AI frontier isn’t the labs anymore - it’s us. It’s figuring out how we build new software that moves us out of the theoretical and into the practical. LLMs are plenty good enough today.
3
3
1
31
1,402
Compute isn't the bottleneck. The missing control loop is. @CoreWeave runs @kubernetesio all the way down. No VMs, just containers and a reconcile loop. That's not a coincidence, it's the only pattern that survives at AI scale. @bassamtabbara explains on Hello, Agent! 👇🎧
The largest inference operations share one pattern: a declarative reconcile loop. @bassamtabbara, Founder and CEO at @upbound_io, points to CoreWeave, running Kubernetes to no VMs, just containers. Click here for the "Hello, Agent!" episode. bit.ly/4gUOAv6 #RealTimeData #DataStreaming #Redpanda
2
4
147
Crossplane v2.4.0 is now available! 🎉 Required resource watches, scale to zero for safe-start providers, vulnerability scannable images, tons of reliability fixes 🛠️ v1.20 will reach end of life (EOL) in November 2026, upgrade to v2 soon! Full details: github.com/crossplane/crossp…
7
1
25
3,070
If claude says "load-bearing" one more time....
8
1
13
1,468
ngl tired of starring at markdown.
63
20
6
731
40,716
phisco retweeted
SWE friend's work right now pretty much amounts to him working from home screaming at his computer like this
19
139
8
5,326
148,886
My mom called me just to tell me she thinks software is more buggy than it's ever been and she is sick of it. She asked me if it was AIs fault. Fellas, we are so back
199
250
44
7,870
186,066
phisco retweeted
I think I finally figured out how to use AI at scale of course, the fact Fable is good is part of it. but I also changed how I work, and it all comes down to one key realization: you don't need to audit the code, but you NEED to audit the *choices* it made. if you just do that, things will work out, and you'll never lose a codebase to chaos. with Fable at least, the following seems to hold: if I give it a good decision: Fable implements it PERFECTLY if I let it decide instead: Fable may make some bad choices that's how I see Fable: as a perfect execution machine capable of converting good decisions into good codebases, no matter how large. given a concrete plan, it lands its implementation. but when anything is underspecified, it can, and will, make bad choices. that's what you must audit. "while working on this, which choices did you make that you're not confident of? list all." then, you just review that. not the git diff, not 1000's of lines of code. just the choices it made along the way. below is a fresh example. overnight, I asked Fable to fix an issue related to MatMul parallelizing worse than expected. it tracked the culprit with perfection, and landed a solution that DID work. but the solution was not general. it just doubled a buffer, which coincidently fixed the program at hands, but the underlying issue was still present. when it completed the job, it declared success. if I just merged it blindly, the issue would still be dormant. that's the main mistake one can do with AI. instead, I asked it to spell out all decisions it made, spotted the bad one, corrected its course, and now the codebase is clean, correct and the issue is gone for good I really think that if you do that religiously - i.e., NEVER merge without this "which decisions you made?" audit - you can go VERY far without ever reading a single line of code. at least on Bend, this is working incredibly well. despite heavy use of AI to implement an ungodly amount of features I could never dream of, the codebase is still in a superb state, with no signs of degradation fresh example below ↓
129
115
37
2,290
289,358
betting on open weights + open source stacks. anything else feels like feudalism. i like market mechanics. i hate feudalism.
29
71
16
941
53,864
phisco retweeted
Control the ideas, not the code: blog post here antirez.com/news/169
63
137
61
1,111
284,085
phisco retweeted
Worth stressing.
18
38
4
558
31,855
Someone DM'd me saying they're using /grill-me in technical interviews Watch the candidate work. See how they answer questions and push back. See whether the AI drives them, or they drive the AI. Honestly genius
81
60
11
2,005
126,776
This. Most tricks quickly become useless with new models, but this approach has proven useful for a few generations already.
Tricks I use when having an agent code review its own work: - tell it to do the code review in a sub-agent *without* the existing session as context. let is read the commit with fresh eyes - tell it to "favor its original judgement" on small issues - use the other top-end model
15
156
yall im ngl its way harder to get joy and satisfaction out of building with ai than it was before constantly straddling burn out, being far less immersed in hard problems, constant context switching i’m tired of the uncertainty of where this is going and how to do it well
276
225
88
3,777
392,754
Today we’re open sourcing Modelplane, the open source control plane for AI inference. Open-weight models changed who runs AI. As inference expands across clouds, accelerators, and regions, operators need a fleet layer above the cluster. That’s @ModelplaneAI. Join us. modelplane.ai/blog/open-cont…
1
4
2
16
2,833
This tbh…
And you all telling me that bash script is a bad idea? you will see more and more of those - and when you perform sudo pacman -Syu you show the same act of trust as when you do bash <url> | curl
1
123
How can the Gmail app suck so much at multi-account management?!? It’s 2026!
39